Compressor Blade Fault Diagnosis Based on Image Processing

An improved method of image analysis is proposed aiming at the problems of high dependence on the inspectors' professionalism during the compressor blades crack detection and the fact that the crack is sometimes blurred and not easy to be noticed. The histogram of a large number of captured images are template matched based on OpenCV, and the images with crack are counted. After that, all the images with crack detected are subjected to erosion and histogram equalization. The crack is amplified by getting a local minimum value. The improved Canny operator is then used to perform edge detection on the image after erosion and histogram equalization to improve the efficiency of crack image segmentation. The USART HMI and STM32 are combined to build a visual fault diagnosis system to judge whether the compressor blade crack length exceeds the allowable value. The system will promptly alarm to remind the maintenance personnel to pay attention to change or repair the blade if the crack size is too large. The simulation results show that the proposed method has good feasibility.